Most internal tools suck. There, I said it.
Even in 2026, if I want to know if a team member is out for a dentist appointment, I usually have to leave Slack, find a bookmark for an HR portal I barely use, log in, navigate a 2010-era calendar UI, and then go back to Slack to finish my thoughts. It’s a 3-minute tax on a 5-second question.
We decided to fix this. At eLEOPARD, we built a POC to see if we could bridge the gap between “Human Language” and “Legacy APIs” using Gemini’s function calling.
Here’s how we did it and why this approach is going to change how we build everything from now on.
The Problem: The "UI Tax"
We noticed that our managers weren’t using the HR tools we paid for because the friction was too high. Instead, they’d ping someone else to ask, “Is John in today?” creating a chain of interruptions.
We didn’t want to build “another dashboard.” We wanted the data to come to the user, where they already live: Slack.
The Tech Stack: Keeping it Lean
We didn’t need a massive infrastructure overhaul. We used:
- Slack SDK: To handle the user input.
- AWS Lambda: Our “stateless” glue.
- Gemini Models: Not just for chat, but for Function Calling.
- Our HRIS API: The source of truth.
How It Works (The "No-Hallucination" Zone)
The biggest fear with LLMs in an enterprise is “hallucination”—we can’t have an AI guessing if someone is on leave.
The breakthrough here is Function Calling. We don’t ask Gemini to tell us who is on leave. We give Gemini a specific “tool” (a function definition) that describes our API.
- The user asks: “Is anyone from the team off next Tuesday?”
- Gemini thinks: “I don’t know the answer, but I see a function called get_team_leaves. I’ll call that for Oct 28th.”
- Lambda executes: It takes that intent, hits our secure HR database, and gets a raw JSON file.
- Gemini synthesizes: It takes that dry data and types back: “Yep, Sarah and Mike are out, but the rest of the team is in.”
This is the key: The AI never sees the database directly, and the database never sees the user’s prompt directly. The Lambda function acts as a secure, audited gatekeeper.
Why This is a Game Changer for Our Clients
Once the “Leave Bot” was live, the realization hit us: We didn’t just build a bot; we built a universal translator for our company’s data. In most companies, data is trapped in silos. To get an answer, you need the right person, the right login, and the right dashboard. By mapping Gemini’s function-calling to your existing APIs, you turn every employee into a “power user” of every system you own.
We’re already seeing how this POC scales into high-friction, high-value areas:
1. Sales & Revenue Operations (The "Pipeline" Pulse)
Instead of a Sales VP chasing a RevOps manager for a report that will be outdated by tomorrow, they ask Slack:
- The Query: “What’s the status of the ACME contract, and who was the last person to touch the account?”
- The Action: Gemini triggers calls to Salesforce (for the stage) and Gong/Slack (for the last interaction).
- The Value: Zero delay in deal intelligence.
2. Engineering & DevOps (Removing the "Context Tax")
Engineers lose hours every week switching between IDEs, Jira, and deployment consoles.
- The Query: “Did the 2 PM production build pass, and if not, what’s the error log summary?”
- The Action: Gemini hits the GitHub Actions or Jenkins API, retrieves the log, and summarizes the failure.
- The Value: Developers stay in the “flow state” longer, accelerating shipping cycles.
3. Real-Time Cloud Governance (FinOps)
Cloud spend is the second or third largest expense for many tech companies, yet visibility is often restricted to a few people with AWS console access.
- The Query: “How much did our SageMaker costs spike yesterday, and was it tied to a specific project?”
- The Action: Gemini queries the AWS Cost Explorer API and maps it against Jira project tags.
- The Value: Immediate fiscal accountability. No more “surprise” $10k bills at the end of the month.
4. Customer Success & Support (The "Proactive" Edge)
Support teams usually have to dig through multiple tickets to understand a customer’s health.
- The Query: “Show me the top 3 open issues for the ‘Big Corp’ account and their current priority.”
- The Action: Gemini pulls data from Zendesk or Freshdesk and presents a summarized briefing.
- The Value: Faster response times and higher CSAT (Customer Satisfaction) scores.
Optimizing Onboarding and Training Costs
Traditional enterprise software often comes with a hidden “Complexity Tax.” Every time a new tool is rolled out, the organization incurs high costs—not just in licensing, but in the hundreds of hours diverted toward training sessions and documentation review.
By transitioning complex workflows into a conversational interface, we significantly lower these barriers. When the “interface” is simply natural language, the need for specialized technical training is drastically reduced. An employee doesn’t need to be an expert in the underlying navigation to get results; they only need to know how to ask a question. This shift allows for faster cross-departmental adoption and ensures that the data within your legacy systems is actually being utilized.
The Executive Takeaway: Architecture is the Strategy
The beauty of this “Function Calling” approach is that it doesn’t require a ‘Rip and Replace’ of your current systems. You keep your secure Salesforce, your proprietary HRIS, and your audited AWS environment. We simply add a conversational “Intelligence Layer” on top. You get the agility of a startup with the security of an enterprise.
The Bottom Line for Leadership
We’re moving away from the era of “training people to use software” and into the era of “software understanding people.”
By using Gemini’s reasoning capabilities as a bridge to our existing APIs, we’ve removed the UI tax. We’re not just saving 3 minutes of a manager’s time; we’re keeping them in “the flow.” That’s where the real ROI is.
Want to Build Something Like This for Your Team?
We help companies turn their existing systems into AI-powered, conversational workflows, without replacing their current stack. Get in touch to explore what this could look like for you.
